Research map / Learning theory and algorithms / Kernels and regression theory
Kernel Approximation via Empirical Orthogonal Decomposition for Unsupervised Feature Learning
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Replace CKN's learned kernel approximation with empirical orthogonal decomposition features | CKN · CIFAR-10 | accuracy: 74.59 -> 76.00 | Helped |
| Replace CKN's learned kernel approximation with empirical orthogonal decomposition features | CKN · CIFAR-100 | accuracy: 43.25 -> 43.66 | Helped |
| Replace CKN's learned kernel approximation with empirical orthogonal decomposition features | CKN · SVHN | accuracy: 91.80 -> 91.98 | Helped |
| Use empirical orthogonal decomposition instead of random features | Hierarchical CKN feature learning · CIFAR-10 | accuracy: 75.93 -> 76.00; 80.73 -> 81.27; 77.68 -> 78.29 | Mixed |
| Use empirical orthogonal decomposition instead of Nyström features | Hierarchical CKN feature learning · SVHN | accuracy: 91.96 -> 91.98; 91.18 -> 91.36; 85.88 -> 86.05 | No clear effect |
| Use empirical orthogonal decomposition instead of Nyström features | Hierarchical CKN feature learning · MNIST | accuracy: 99.38 -> 99.36; 99.48 -> 99.43; 99.47 -> 99.51 | Mixed |
| Reduce the number of final-layer feature maps (ablation) | CKN, Random, Nyström, and Proposed · CIFAR-10 | accuracy | Mixed |
| Use a Gaussian mixture rather than a single Gaussian for the input distribution (ablation) | Proposed kernel approximation · Synthetic data | normalized spectral-norm approximation error | Hurt |
Closest papers by meaning
- Fast and Accurate Refined Nyström-Based Kernel SVM (AAAI 2016 · similarity 0.88)
- Approximate Kernel Selection with Strong Approximate Consistency (AAAI 2019 · similarity 0.87)
- Learning Kernels with Random Features (NeurIPS 2016 · similarity 0.87)
- Quadrature-based features for kernel approximation (NeurIPS 2018 · similarity 0.87)
- Learning Adaptive Random Features (AAAI 2019 · similarity 0.86)
- Learning in Feature Spaces via Coupled Covariances: Asymmetric Kernel SVD and Nyström method (ICML 2024 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.